dracor-org / ezlinavis Goto Github PK

In order to update our dependencies and bring ezlinavis more in line with the dracor-frontend we should migrate from react-bootstrap to reactstrap which depends on bootstrap 4.

Value in 4th column of CSV should be used to put weight on edges (make lines thicker, according to value).

along those lines:

"Easy Linavis (ezlinavis) generates CSV files with network data from simple segmentations of dramatic texts. In the left column, you can list segments (chapters, acts, scenes, etc.) and characters appearing or speaking in a given segment. Segments are indicated with a hashtag and they can be hierarchical (e.g., # First Act / ## First Scene / ### …). This will automatically generate a CSV file with node-node relations (source, type, target, weight) in the column in the centre. Data changes as you type: as soon as you change something in the first column, the mid-column changes accordingly. The "type" column in the CSV file is always "undirected" here, but we inserted it so you can directly work with the CSV files in Gephi. The network graph in the right column is also generated live, using a spring-embedded layout, just to give you a first impression of what your network data looks like. To make it easier to understand how ezlinavis works, we provide some example files which can be accessed via the drop-down menu in the first column.

ezlinavis was developed by Carsten Milling and Frank Fischer, making use of the React and Sigma JS libraries. It is mainly meant for didactic purposes, although it could also be suitable to process simply-structured research data. We are mainly resorting to it in our workshops on network analysis of literary texts."

When the text input uses CRLF (\r\n) line endings the parser takes very long (with just a few lines) or does not return at all. As a workaround example files should always use linefeeds (\n) only. For a proper solution the grammar should be modified to prevent the parser from stalling.

See dracor-org/dracor-frontend#275.

I see two problems when starting ezlinavis in Firefox:

1. The graph seems to be placed in the vertical centre of the CSV column. If this list is long, the graph doesn't show before scrolling down (maybe it's enough to top-align the graph?).
2. The graph-layout algorithm seems to work differently in Firefox, if you compare the graphs for "Faust, pt. 1" in Firefox and Chrome. The latter, although always a bit differently of course, usually shows nice clusters around Faust. In Firefox, the graph looks like a haystack. Why could the two browsers show different results?

1. At the moment, having "Name" and "Name " (the latter with trailing whitespace) in two lines yield two different nodes. This has been reported as confusing by some, so could we ignore trailing spaces (and only them) when extracting node labels?
2. For the delimiter, we need a hashtag followed by a whitespace at the moment. Wouldn't it be enough to have just the hashtag without the whitespace? This would shorten the explanation how to delimit co-occurrences.

create-react-app provides a more modern development environment and ezlinavis (now hosted on its own domain) could profit from the service worker setup that comes with it.

… via gh-pages.